FC
EA
15 August 2016
Preparing the data
### Notes
# install.packages('ggplot2',dependencies = T)
# Test code line with command+enter
# update.packages(checkBuilt = T,ask = F)
# head(ds) # first few lines of a dataframe
# str(ds) #show structure of dataframe
###
# Setting variables:
options(width=108)
outT <- 2.5 # Outlier factor
# Loading libraries
library(ggplot2) # for plotting
library(plyr)
#Importing the data
#data_dir <- '/Volumes/PK/'
#homeDir <- path.expand('~/..')
homeDir <- path.expand('~')
dataDir <- paste(homeDir, 'Dropbox/Projects/fc/fc/data/usable/extracted', sep='/')
dataFiles <- dir(dataDir, pattern='csv')
nSubj <- length(dataFiles) # number of subjects
# Going through all subjects' data:
df <- data.frame() # data frame that combines all subjects' data into a single ds
#curSubjN <- 6 #temp
for(curSubjN in 1:nSubj){
# Loading the data;
ds <- read.csv(paste(dataDir, dataFiles[curSubjN], sep='/'))
# Shorter names for variables:
colnames(ds) <- c('subjId', 'domEyeR', 'threshStHi', 'threshStLo', 'thresh', 'trialN', 'sentId',
'sentPx', 'congr', 'fam', 'locTop', 'cued', 'corr', 'broken', 'st')
# Non-blank data set:
cds <- ds[ds$sentId<31,]
# Removing the outliers:
outlLo <- cds$st>(mean(cds$st)-outT*sd(cds$st)) # binary for every row
outlHi <- cds$st<(mean(cds$st)+outT*sd(cds$st))
cds$outlNlo <- sum(!outlLo) # counting the number of outliers before removing
cds$outlNhi <- sum(!outlHi)
cds <- cds[outlLo,] # removing the outliers
cds <- cds[outlHi,]
cds <- cds[!is.na(cds$subjId),] # usually don't need this; remove NA rows
# Normalized suppression times:
cds$stNorm <- cds$st / mean(cds$st)
# Binding the subject ds to the common data frame:
df <- rbind(df, cds)
}
# Prettying up the data frame:
df$subjId <- factor(df$subjId)
df$Familiarity[df$fam==1] <- 'Familiar\n(Upright)'
df$Familiarity[df$fam==0] <- 'Unfamiliar\n(Inverted)'
df$Congruency[df$congr==1] <- 'Congruent'
df$Congruency[df$congr==0] <- 'Incongruent'
df$Attention[df$cued==1] <- 'Cued'
df$Attention[df$cued==0] <- 'Uncued'
# Binning
df$bin <- 6
df$bin[df$trialN<500] <- 5
df$bin[df$trialN<400] <- 4
df$bin[df$trialN<300] <- 3
df$bin[df$trialN<200] <- 2
df$bin[df$trialN<100] <- 1
head(df)## subjId domEyeR threshStHi threshStLo thresh trialN sentId sentPx congr fam locTop cued corr broken
## 1 14 1 0.31181 0.23408 0.27294 1 5 133 1 0 1 1 1 1
## 2 14 1 0.31181 0.23408 0.27294 2 28 108 1 0 0 0 1 1
## 3 14 1 0.31181 0.23408 0.27294 5 27 104 0 0 1 1 1 1
## 4 14 1 0.31181 0.23408 0.27294 6 21 109 0 0 1 1 1 1
## 5 14 1 0.31181 0.23408 0.27294 8 18 117 0 0 1 1 1 1
## 6 14 1 0.31181 0.23408 0.27294 9 10 128 0 0 0 0 1 1
## st outlNlo outlNhi stNorm Familiarity Congruency Attention bin
## 1 1.4901 0 7 1.0650477 Unfamiliar\n(Inverted) Congruent Cued 1
## 2 1.2797 0 7 0.9146645 Unfamiliar\n(Inverted) Congruent Uncued 1
## 3 1.5523 0 7 1.1095051 Unfamiliar\n(Inverted) Incongruent Cued 1
## 4 1.4809 0 7 1.0584720 Unfamiliar\n(Inverted) Incongruent Cued 1
## 5 1.3432 0 7 0.9600510 Unfamiliar\n(Inverted) Incongruent Cued 1
## 6 1.1362 0 7 0.8120980 Unfamiliar\n(Inverted) Incongruent Uncued 1Quality control
Summary stats per participant
ddply(df, c('subjId'), summarise, outlNumLow = median(outlNlo), outlNumHigh = median(outlNhi),
cuedCaught = sum(cued & corr), uncuedCaught = sum(!cued & corr))## subjId outlNumLow outlNumHigh cuedCaught uncuedCaught
## 1 14 0 7 116 116
## 2 17 0 9 115 116
## 3 19 0 10 113 107
## 4 21 0 6 118 116
## 5 23 0 6 117 116
## 6 24 0 0 105 101
## 7 25 0 9 115 116
## 8 26 0 9 118 113
## 9 27 0 11 110 115
## 10 28 1 7 115 116
## 11 29 0 14 114 112
## 12 30 0 5 118 117
## 13 31 0 7 116 115
## 14 32 0 7 119 114
## 15 33 0 11 107 104
## 16 34 0 9 113 117
## 17 36 0 6 118 115
## 18 37 0 4 119 117
## 19 38 1 4 115 116
## 20 39 0 7 117 116
## 21 41 0 7 119 114
## 22 42 0 6 120 114
## 23 43 4 4 115 116
## 24 44 0 6 118 111sts across trials
for(curSubjN in 1:nSubj){
curSubjId <- levels(df$subjId)[curSubjN]
print(curSubjId)
ss <- df[df$subjId==curSubjId,]
p <- ggplot(data=ss, aes(x=trialN, y=st)) + geom_line() + theme_bw() + ylim(.5,2) +
labs(x='Trial Number', y='Suppression Time', title=paste('Participant', as.character(curSubjId))) +
theme(panel.grid.minor.x=element_blank(), panel.grid.minor.y=element_blank(),
axis.text=element_text(size=8), axis.title=element_text(size=9),
legend.text=element_text(size=8), legend.title=element_text(size=9),
legend.key = element_blank(), #legend.margin=unit(-.04, 'in'),
legend.background = element_rect(fill='transparent'))
plot(p)
}## [1] "14"
## [1] "17"
## [1] "19"
## [1] "21"## Warning: Removed 1 rows containing missing values (geom_path).
## [1] "23"## Warning: Removed 1 rows containing missing values (geom_path).
## [1] "24"
## [1] "25"
## [1] "26"
## [1] "27"
## [1] "28"
## [1] "29"## Warning: Removed 1 rows containing missing values (geom_path).
## [1] "30"
## [1] "31"
## [1] "32"
## [1] "33"
## [1] "34"
## [1] "36"
## [1] "37"
## [1] "38"
## [1] "39"## Warning: Removed 2 rows containing missing values (geom_path).
## [1] "41"## Warning: Removed 2 rows containing missing values (geom_path).
## [1] "42"
## [1] "43"## Warning: Removed 1 rows containing missing values (geom_path).
## [1] "44"
Individual plots
Congruence X Familiarity X Cue
for(curSubjN in 1:nSubj){
curSubjId <- levels(df$subjId)[curSubjN]
ss <- df[df$subjId==curSubjId,]
p <- ggplot(data=ss, aes(x=factor(Familiarity), y=st, colour=factor(Congruency))) + geom_boxplot() +
facet_grid(~Attention) + theme_bw() +
labs(x='Familiarity', y='Suppression Time', colour='Congruency',
title=paste('Participant', as.character(curSubjId))) +
theme(panel.grid.minor.x=element_blank(), panel.grid.minor.y=element_blank(),
axis.text=element_text(size=8), axis.title=element_text(size=9),
legend.text=element_text(size=8), legend.title=element_text(size=9),
legend.key = element_blank(), #legend.margin=unit(-.04, 'in'),
legend.background = element_rect(fill='transparent'))
plot(p)
}
Group plots
Is cueing effective across participants?
p <- ggplot(data=df, aes(x=Attention, y=st)) + geom_boxplot() + facet_grid(~subjId) + theme_bw() +
labs(x='Attention effect X Subject', y='Suppression Time', colour='Attention',
title='Effect of attentional cueing') +
theme(panel.grid.minor.x=element_blank(), panel.grid.minor.y=element_blank(),
axis.text=element_text(size=8), axis.title=element_text(size=9),
legend.text=element_text(size=8), legend.title=element_text(size=9),
legend.key = element_blank(), #legend.margin=unit(.0, 'in'),
legend.background = element_rect(fill='transparent'))
plot(p)
Is cueing effective across trials?
p <- ggplot(data=df, aes(x=Attention, y=st)) + geom_boxplot() + facet_grid(~bin) + theme_bw() +
labs(x='Attention effect X 100-trial bins', y='Suppression Time', colour='Attention',
title='Effect of attentional cueing') +
theme(panel.grid.minor.x=element_blank(), panel.grid.minor.y=element_blank(),
axis.text=element_text(size=8), axis.title=element_text(size=9),
legend.text=element_text(size=8), legend.title=element_text(size=9),
legend.key = element_blank(), #legend.margin=unit(.0, 'in'),
legend.background = element_rect(fill='transparent'))
plot(p)
Congruence X Familiarity X Cue
dfSum <- ddply(df, .(subjId, Attention, Familiarity, Congruency), summarise,
`Mean Normalized ST` = mean(stNorm, na.rm=T))
p <- ggplot(data=dfSum, aes(x=factor(Familiarity), y=`Mean Normalized ST`,
colour=factor(Congruency))) +
geom_boxplot() + facet_grid(~Attention) + theme_bw() +
labs(x='Familiarity', y='Suppression Time', colour='Congruency',
title='Congruence X Familiarty X Cue') +
theme(panel.grid.minor.x=element_blank(), panel.grid.minor.y=element_blank(),
axis.text=element_text(size=8), axis.title=element_text(size=9),
legend.text=element_text(size=8), legend.title=element_text(size=9),
legend.key = element_blank(), #legend.margin=unit(-.04, 'in'),
legend.background = element_rect(fill='transparent'))
plot(p)
p <- ggplot(data=dfSum, aes(x=factor(Familiarity), y=`Mean Normalized ST`,
group=factor(subjId), colour=factor(subjId))) +
geom_line() + facet_grid(~Attention*Congruency) + theme_bw() +
labs(x='Familiarity', y='Suppression Time', colour='Congruency',
title='Congruence X Familiarty X Cue') +
theme(panel.grid.minor.x=element_blank(), panel.grid.minor.y=element_blank(),
axis.text=element_text(size=8), axis.title=element_text(size=9),
legend.text=element_text(size=8), legend.title=element_text(size=9),
legend.key = element_blank(), #legend.margin=unit(-.04, 'in'),
legend.background = element_rect(fill='transparent'))
plot(p)
Linear models
Centered data set
df$congrC <- -1
df$congrC[df$congr==1] <- 1
df$famC <- -1
df$famC[df$fam==1] <- 1
df$cuedC <- -1
df$cuedC[df$cued==1] <- 1
df$topSubj <- 0
df$topSubj[df$locTop==df$cued] <- 1Mixed linear regressions
#mFull <- lmer(st ~ famC * cuedC * congrC + (1|subjId), df)
library(BayesFactor)## Loading required package: coda## Loading required package: Matrix## ************
## Welcome to BayesFactor 0.9.12-2. If you have questions, please contact Richard Morey (richarddmorey@gmail.com).
##
## Type BFManual() to open the manual.
## ************df$subjId <- as.factor(df$subjId)
df$fam<- as.factor(df$fam)
df$cued<- as.factor(df$cued)
df$congr<- as.factor(df$congr)
bf <- anovaBF(st ~ fam * cued * congr + subjId, data=df, whichRandom = 'subjId')
bf## Bayes factor analysis
## --------------
## [1] congr + subjId : 0.03315989 <U+00B1>1.1%
## [2] fam + subjId : 0.03010921 <U+00B1>1.38%
## [3] congr + fam + subjId : 0.001010881 <U+00B1>5.28%
## [4] congr + fam + congr:fam + subjId : 4.809882e-05 <U+00B1>2.07%
## [5] cued + subjId : 3420866 <U+00B1>1.13%
## [6] congr + cued + subjId : 111867.5 <U+00B1>1.5%
## [7] fam + cued + subjId : 112615.9 <U+00B1>8.59%
## [8] congr + fam + cued + subjId : 3469.594 <U+00B1>3.03%
## [9] congr + fam + congr:fam + cued + subjId : 283.9225 <U+00B1>42.6%
## [10] congr + cued + congr:cued + subjId : 13291.54 <U+00B1>3.91%
## [11] congr + fam + cued + congr:cued + subjId : 363.0126 <U+00B1>2.27%
## [12] congr + fam + congr:fam + cued + congr:cued + subjId : 18.15136 <U+00B1>2.93%
## [13] fam + cued + fam:cued + subjId : 5456.2 <U+00B1>17.67%
## [14] congr + fam + cued + fam:cued + subjId : 163.059 <U+00B1>14.97%
## [15] congr + fam + congr:fam + cued + fam:cued + subjId : 7.474029 <U+00B1>10.06%
## [16] congr + fam + cued + congr:cued + fam:cued + subjId : 16.63911 <U+00B1>5.61%
## [17] congr + fam + congr:fam + cued + congr:cued + fam:cued + subjId : 0.8475175 <U+00B1>5.38%
## [18] congr + fam + congr:fam + cued + congr:cued + fam:cued + congr:fam:cued + subjId : 0.08806365 <U+00B1>11.1%
##
## Against denominator:
## st ~ subjId
## ---
## Bayes factor type: BFlinearModel, JZS